Linux on Power: Where It Can Run in the Cloud
Should Linux stay on Power, or move to x86?
Linux is the portable member of the Power family. Unlike AIX and IBM i, it runs on both Power and x86, so this is a strategy question rather than a feasibility question. Keep it on cloud Power (IBM Power Virtual Server, a managed or hosted Power cloud, or Skytap on Azure) when the Linux tier sits beside AIX or IBM i you are keeping on Power, when latency between those tiers matters, or when one operating model is worth more than a bigger ecosystem. Replatform it to x86 when your tooling, skills, and ISV support all point that way and nothing on Power is holding it in place. The deciding input is usually the ISV support matrix, not the infrastructure.
At a glance
Key takeaways
Portability changes the question
AIX and IBM i give you one realistic platform. Linux gives you two. That means the Linux tier of a Power estate should be decided on strategy and cost, not on what is technically possible.
Staying on Power buys one operating model
Keeping Linux beside AIX or IBM i in the same cloud Power platform means one provider, one region, one set of runbooks, and no latency penalty between tiers. That is worth real money in operational overhead.
The ISV matrix usually decides it
Current Linux on Power is little-endian ppc64le. Open-source and container workloads are generally well served. Commercial ISVs vary, and one x86-only application in the tier can settle the whole question.
A move to x86 is a rebuild, not a copy
Far easier than an AIX or IBM i migration, because the operating system is portable. Still a real project: compiled code needs rebuilding for the target architecture, and every workload needs its own test cycle.
Why Linux Is the Different One
Most of the hard decisions in an IBM Power estate come from lock-in. IBM i runs on Power and nowhere else. AIX runs on Power and nowhere else. When those workloads move, the realistic destinations are all still Power: your own hardware, IBM Power Virtual Server, or a partner-run Power cloud. The debate is about who operates the hardware, not about which architecture you land on.
Linux breaks that pattern. It runs on Power and on x86, so a Linux tier inside a Power estate has two genuinely different futures available to it. You can keep it on cloud Power, sitting beside the AIX and IBM i systems it probably talks to, or you can treat it like any other Linux workload and consolidate it onto x86 with the rest of your fleet.
One technical detail matters for planning. Current Linux distributions on Power are little-endian (the ppc64le architecture), which aligns Linux on Power with the x86 software ecosystem and is why moving between them is a rebuild rather than a rewrite. Red Hat Enterprise Linux, SUSE Linux Enterprise Server, and Ubuntu Server all ship Power builds, though you should confirm the exact supported versions with both the distribution vendor and your chosen Power cloud provider.
If your question is the reverse one, whether to move AIX workloads onto Linux in the first place, that is a different decision with different economics. Our comparison of staying on AIX versus moving to Linux covers the endianness, toolchain, and ISV issues that drive it.
Where Linux on Power Can Run
Four realistic destinations, three of which keep you on Power. Supported distributions and versions vary by platform and region, so confirm the current matrix with the provider before you plan around it.
Best fit
IBM Power Virtual Server
IBM-operated Power infrastructure running AIX, IBM i, and Linux LPARs as a consumption-billed cloud service. The natural fit when you want IBM to own the hardware lifecycle and you are already moving other Power workloads there.
Best fit
Managed or hosted Power cloud
A partner runs the Power platform and takes on more of the operational load: monitoring, patching, backup, and often DR. The fit when the attraction is offloading work rather than just offloading hardware.
Best fit
Skytap on Azure
Power workloads running inside an Azure region, which keeps the Linux tier close to Azure-native services while staying on Power. Worth evaluating when the wider strategy is already Azure-centric.
Key decision
Replatform to x86 Linux
The only destination that leaves Power behind. Opens the full hyperscaler market, pricing, and skills pool, at the cost of a rebuild-and-test project per workload and a split estate until the Power side is retired.
The Options Compared
| Option | Migration effort | Ecosystem and skills | Operating model | Best when |
|---|---|---|---|---|
| Stay on cloud Power (PowerVS) | Low: an LPAR move, no architecture change | Smaller ppc64le ecosystem; existing Power skills carry over | One platform for AIX, IBM i, and Linux | Other Power workloads are moving to PowerVS anyway |
| Stay on Power, managed or hosted | Low: an LPAR move, plus provider onboarding | Smaller ecosystem; the provider supplies scarce Power skills | One platform, with operations largely outsourced | Thin platform staffing is the real problem to solve |
| Stay on Power via Skytap on Azure | Low to moderate: LPAR move into an Azure region | Power ecosystem, adjacent to Azure-native services | Power estate inside an Azure landing zone | The corporate strategy is already Azure-first |
| Replatform to x86 Linux | Moderate: rebuild and test per workload, no OS rewrite | Full mainstream ecosystem, tooling, and hiring pool | Two platforms until the Power tier is fully retired | Nothing on Power holds it, and ISVs support x86 only |
Notice what is not in the table: feasibility. Every row works. That is the point, and it is why Linux decisions inside a Power estate get made badly. With AIX and IBM i, the constraints narrow the field for you. With Linux, nothing narrows it, so whoever is in the room with the strongest platform preference tends to win the argument. Force the decision back onto integration, ISV support, cost, and skills.
How to Decide
- 1
Count what you actually run
Inventory the Linux on Power estate: LPARs, distributions and versions, applications, and cores. A handful of utility LPARs and a large production Linux tier deserve different answers.
- 2
Check the ISV support matrix first
For every commercial application in the tier, confirm whether the vendor supports ppc64le. One x86-only dependency can settle the question before any cost modeling starts.
- 3
Map the integration with AIX and IBM i
If the Linux tier is chatty with Power-resident systems you are keeping, splitting it onto x86 adds network hops and a latency budget that has to be engineered and tested.
- 4
Price both on equal terms
Compare fully loaded run-rates over the same horizon: Power cores and memory against x86 instances, plus the one-time rebuild-and-test cost on the x86 side that sticker comparisons omit.
- 5
Cost the operating model, not just the compute
Running one platform is cheaper than running two: one provider relationship, one set of runbooks, one on-call rotation. Put a number on that before you decide to split the estate.
- 6
Decide per workload, not estate-wide
Some Linux workloads belong beside the Power estate and some belong on x86. A split answer is a legitimate outcome, provided each side was chosen rather than inherited.
What a Move to x86 Actually Involves
Compared with an AIX or IBM i migration, this is the easy one. The operating system is the same family on both sides, the administration model transfers, and there is no platform to rewrite underneath the applications. That is genuinely different from a modernization program on IBM i, where the application itself is the project.
It is still a real migration. Workloads sort into three groups. Java, scripted, interpreted, and containerized applications generally move with modest effort, provided container images are built for the target architecture. Compiled C and C++ needs rebuilding for the destination and a proper regression cycle. Anything holding vendor binaries with no x86 equivalent, or code with architecture assumptions baked in, is the expensive tail that decides the schedule.
Plan it like any other migration: inventory, pilot one representative workload, then move the rest in waves with testing gates. Our IBM Power migration timeline guide covers the phase structure and the items that usually stretch the calendar.
Common Mistakes
Use caution
Letting a platform preference decide it
Because every option is feasible, the loudest opinion in the room often wins. A recommendation should follow from your integration map, ISV matrix, and cost model, not from what a partner happens to sell.
Use caution
Assuming the x86 port is free
It is cheaper than an AIX move, not free. Rebuilds, container image rework, regression testing, and runbook changes are all real hours. Get that number on the table before comparing monthly rates.
Use caution
Ignoring latency to Power-resident tiers
If the Linux tier talks constantly to AIX or IBM i systems staying on Power, splitting the estate introduces a network path where there used to be a backplane. Measure it in the pilot, not after cutover.
Use caution
Splitting the estate without costing the operating model
Two platforms means two provider relationships, two sets of runbooks, and two operational skill sets. Shops that skip this arithmetic often find the savings on compute were smaller than the added overhead.
Frequently Asked Questions
Is Linux on Power locked to Power hardware?
No, and that is what makes it different from AIX and IBM i. Linux runs on both Power and x86, so you can keep it on cloud Power beside the rest of your estate or replatform it to x86 and the hyperscalers. The decision is about consolidation strategy and cost, not about feasibility. AIX and IBM i do not offer that choice.
Where can Linux on Power run in the cloud?
On IBM Power Virtual Server, on a managed or hosted Power cloud from a partner, or on Skytap on Azure, all of which run Linux LPARs on IBM-operated or partner-operated Power hardware. If you replatform to x86 Linux instead, it can run on any hyperscaler or hosting provider. Verify the current supported distribution and version matrix with each provider before committing, because it changes.
How hard is it to move Linux from Power to x86?
It is a real migration with per-workload testing, but far simpler than an AIX or IBM i move, because the operating system itself is portable and you are not rewriting the platform underneath. Java, scripted, containerized, and interpreted workloads usually move with modest effort. Compiled C and C++ needs a rebuild for the target architecture, and anything with byte-order assumptions or vendor binaries needs real attention.
Which Linux distributions run on Power?
Red Hat Enterprise Linux, SUSE Linux Enterprise Server, and Ubuntu Server all ship builds for Power in the little-endian ppc64le architecture, which is what current Linux on Power means. Confirm the specific version and support terms with both the distribution vendor and your Power cloud provider, since supported combinations vary by platform and by region.
Should I keep Linux on Power or move it to x86?
It depends on three things: how tightly the Linux tier integrates with AIX or IBM i systems you are keeping on Power, what your ISVs support on ppc64le, and where your team's skills and tooling already sit. Staying on Power gives you one platform, one region, and one operating model. Moving to x86 opens the full cloud-native market but splits your estate across two architectures until the Power side is gone.
Will my software vendors support Linux on Power?
Some will and some will not, and this is the single most common thing that decides the answer. Major infrastructure software, open-source stacks, and container workloads are generally well served on ppc64le, but plenty of commercial ISVs ship x86 builds only. Check the support matrix for every application in the Linux tier before you plan around staying on Power. A vendor that does not certify ppc64le makes the decision for you.
Sources
- Confirm the supported Linux distribution and version matrix with each Power cloud provider; supported combinations vary by platform and region.
- Confirm ppc64le support for every commercial application in the Linux tier with the software vendor directly.
- Verify current pricing and instance ceilings with IBM or the relevant provider before modeling a comparison.
Related pages
Place Your Linux on Power Workloads Deliberately
We weigh staying on cloud Power against an x86 move on integration, ISV support, cost, and skills, and tell you when a split estate is the right answer.